Comparison of four measurement schedules
for determination of soil particle-size
distribution by the hydrometer method
P.G. KARKANIS1, K. AU2 and G.B. SCHAALJE3
Canadian Challenger AssociatesLtd., Box 1561, Lethbridge, AB, Canada T1J4K3;2Land Evaluation andReclamation Branch,
Irrigation andResource Management Division, AlbertaAgriculture, Agriculture Centre, Lethbridge, AB,Canada TIJ 4C7; and
Scientific Support Unit, Agriculture Canada, Agriculture Centre, Lethbridge, AB, Canada TIJ 4B1. Received 4 July 1990;
accepted 15 February 1991.
Karkanis, P.G., Au, K. and Schaalje, G.B. 1991. Comparison of four
measurement schedules for determination of soil particle-size dis
tribution by the hydrometer method. Can. Agric. Eng. 33:211-215.
Four schedules of measurement by the hydrometer method were ex
amined for determining the particle-size distribution of soils in
southern Alberta. All methods utilized the same sample pretreatment
procedures and differed only in the time at which hydrometer readings
were taken. Linear and quadratic regression analyses were used to
describe the relationship between estimates for sand and clay content
determined by the different methods. Comparisons among die meth
ods were made by testing the hypothesis that the relationship between
any two methods was linear and had a slope of one. Clay content was
significantly overestimated with the Bouyoucos (1962) method when
compared with the other three methods. Regression equations were
developed for conversion of sand and clay content, as determined by
the Bouyoucos (1962)method, to the more theoretically soundmeth
ods of Day (1965), USDI (1982), and Gee and Bauder (1986).
KeyWords:Particle-size distribution, hydrometer method, soil tex
agricultural soils. Sieving methods are only suitable for sepa
ration of the sand fraction of soils, i.e. particles larger than
0.05 mm diameter (Baver et al. 1972). The pipette and hy
drometer methods represent two approaches to the
determination of PSD by the sedimentation method. The pi
pette method is generally considered more accurate than the
hydrometer method (Day 1965), however, comparableresults
can be obtained provided similar pretreatment techniques are
employed (Gee and Bauder 1986).
Sedimentation methods are based on the relationship be
tween the settling velocity and diameter of a spherical particle
in a fluid at constant temperature, described by Stokes' law
(Baver et al. 1972). The strict applicability of Stokes' law to
PSD determinations has been the subject of much debate due
to variations in the shape and density of clay-sized particles,
however, for most practical purposesa reasonable estimateof
Quatreseries de mesureshydrometriques ont ete examinees afin de
determiner la distribution granulometrique de sols dans le sud de
FAlberta. Toutes les methodes utilisaient les memes procedures de
particle size canbe obtained (Baver et al. 1972). Particle size
may alsobe defined in different ways on the basis of several
arbitrary criteria that apply to spherical particles but do not
necessarily holdtrue for the anisometric particles occurring in
pretraitement d'echantillons et ne differaient queparle moment ou les
the soil (Day 1965).
ture, soil analysis.
mesures avaient ete prises. Par des analyses de regression lineaire et
quadratique, on a pu determiner les relations entreles estimations de
teneur en sable et en argileetablies par les differentes methodes. Des
comparaisons ont ete fakes en se fondant sur rhypothese que les
relationsentre deux methodes, quelles qu'elles soient, etaient lineaires
et fondees sur une pente de un. La teneur en argile etait
considerablement surestimee par la methode Bouyoucos (1962), com-
parativement aux trois autres. Des equations de regression furent
etablies pour convertir la teneuren sableet en argiledeterminee par la
methode Bouyoucos(1962), suivant les methodes theoriquement plus
suresde Day (1965),d'USDI (1982), et de Gee et Bauder (1986).
INTRODUCTION
Particle-size distribution (PSD) of the fine earth (less than 2
mm diameter) fraction is an important measurement for char
acterization of soils for irrigation and drainage purposes. Soil
texture is used in evaluating irrigation suitability, assessing
land drainability, predicting hydraulic conductivity and esti
mating moisture-retention characteristics (Alberta Agriculture
1983).
Several methods are available for measurement of PSD, but
sieving and sedimentation techniques are generally used for
CANADIAN AGRICULTURAL ENGINEERING
The Bouyoucos (1962) hydrometer method is commonly
used for determination of PSD in agricultural soils in Alberta.
This method uses a 40 s hydrometer reading to estimate the
sandcontent and a 2 h hydrometer reading to approximatethe
clay fraction. The 2 h hydrometer reading yieldsanestimate of
the less than 0.005 mm silt and clay fraction rather than the less
than0.002 mm clay fraction. When more accurate differentia
tion between silt and clay is required, the Bouyoucos method
is not recommended (Gee and Bauder 1986).
Day (1965), USDI (1982), and Gee and Bauder (1986)
describe alternativeapproaches that overcome the majorcriti
cism of the Bouyoucos method. The Day (1965) method
involves hydrometer readings at 0.5,1,3,10,30,90,270 and
720 min that are plotted against calculated particle diameters
to obtain estimates of PSD through graphical interpolation.
Sand content is estimated from a 40 s hydrometer reading and
clay content from a 8 h reading in the USDI (1982) method.
Hydrometer readings are obtained at 30 and 60 s for determi
nation of sand content and at 1.5 and 24 h for calculation of
clay content in the simplified Gee and Bauder (1986) method.
The purpose of this study was to evaluate four hydrometer
211
methods for determining the PSD of southern Alberta soils
from a wide range of textural classes. Relationships between
estimates of PSD determined at specific settling times in the
different methods were also described.
MATERIALS AND METHODS
A total of 177 samples from mineral horizonsof soils from the
Brown and Dark Brown soil zones in southern Alberta were
used for this study. These soils were generally low in organic
matter (less than 1%) and represented a wide range of textural
classes from sand to heavy clay (Fig. 1). Sand content ranged
from 6 to 95 percent and clay content varied from 3 to 82
cylinder containing the soil suspension was then placedon a
level surface ina room having a constant temperature of22°C.
A stop watch was used for timing measurements of the soil
suspension at sedimentation times of 30,40,50 and 60 s and
at 1.5,2,4,6,8 and 24 h. The hydrometer was immersed in the
suspension 10 s before each reading and was removed imme
diately following each reading, except for the first four
measurements. Density measurements were also obtained at
the same time intervals on a blank solution of sodium-
hexametaphosphate made up to the same 1130 mL volume. A
corrected hydrometer reading was obtained at each settling
time as:
percent
C = R\-Ri
(1)
where:
C = mass of soil in suspension at time hydrometer reading
was taken (g»L-i\),
R\ = density of soil suspension (g*L-i\), and
Rl =density ofblank solution (g«L-1).
Sand, silt, and clay contents were then calculated for all
samples according to the settling times used in the Bouyoucos
(1962), Day (1965), USDI (1982) and simplified Gee and
Bauder (1986) methods. Calculations were based on air-dry
weight of soil to conform with procedures used in routine PSD
analysis in most agricultural soil laboratories in Alberta.
Linear and quadratic regression analyses were used to ex
20
40
60
80
100
Sand content, wt %
Fig. 1. Texture of soil samples used according to the Day
(1965) method.
amine relationships among the methods. Nonlinearity of the
relationships was examined by testing the hypothesis that the
coefficient of the seconddegreeterm wasequal to zero. When
a relationship between two methods waslinear, agreement of
the methods was examined by testing the hypothesis that the
intercept wasequal to zeroand the slopewasequalto one.In
all cases the coefficient of determination (R2) was used as a
measure of the goodness of fit between two methods.
RESULTS AND DISCUSSION
All soil samples were subjected to the same pretreatment
and dispersion procedures. A50g sample ofthefine earth (less
than 2mm diameter) fraction ofthe soils was soaked overnight
two ofthemethods formeasuring sand content (Table I).Inall
dissolved in distilled water to enhance dispersion (Gee and
cases the slopes were significantly different from one. All of
the methods were thus systematically different from each
in abeaker containing 50 g-L"1 ofsodium-hexametaphosphate
Bauder1986). Otherpretreatmentprocedures suchas removal
of carbonates, solublesalts, organicmatteror iron oxideswere
not undertaken.
Each sample was subsequently dispersed mechanically by
A significant linear relationship was observed between any
other, but the results from any one method would be very
accurately predicted from the results of any other method
using the regression equations (Table I).
mixing cup and was stirred with anelectric mixer at 1500 rpm
In spite of a non-zero intercept and a slope significantly
different from one,it canbe seenthattherelationship between
sand content determined according to Geeand Bauder (1986)
for 10 min. The mixed suspension was then transferred to a
and Bouyoucos (1962) or USDI (1982) was close to 1:1 in all
transferring the soil suspension from the beaker to a 1 L
1130 mL hydrometer cylinder (Bouyoucos 1962) and made up
textural classes (Fig. 2). In contrast, the relationship between
to volume by adding distilled water with the density hydrom
eter in suspension. The density hydrometer was then removed
and the suspension was left overnight to equilibrate to room
temperature. Prior to commencing the tests the next day, the
hydrometer wasagainplacedin thesuspension andthevolume
was remade up to 1130 mL. An ASTM No. 152H density
sand content determined according to Day (1965) and that
determined by both other methods (Gee and Bauder 1986,
Bouyoucos1962,USDI 1982)was close to 1:1 only for soils
with less than 40% sand. For soils with more than 40% sand,
the Day (1965) method gave a higher estimate of sand
content than the other methods. This may be due to the use
of graphicalvalues in determiningresults in the Day (1965)
method, rather than the use of a formula in determining
hydrometer was used for all measurements.
Particle-size distribution tests were started byremoving the
hydrometer from the glass cylinder and mixing the soil suspen
sion by turning the cylinder upside-down twenty times. The
212
results, as in the other methods.
Gee and Bauder (1979) found that the difference between
KARKANIS, AU and SCHAALJE
Table I. Comparison of percentsand content measured by four hydrometer methods.
Regression equation
Method
SEof
SEof 1st deg.
SEof
intercept
coefficient
prediction
Y2 = 0.666 +0.995 (Yi)
0.081t
0.002*
0.999
0.60
Day vs Bouyoucos
Day vs USDI
Y2=-0.184+1.039 (Yi)
0.174
0.004*
0.997
1.30
Day us Gee & Bauder
Y2 =-0.870+1.044 (Yi)
0.171
0.004*
0.997
1.27
ttGee & Bauder vs Bouyoucos
Gee & Bauder vs USDI
t y-interceptsignificantlydifferent from zero, p<0.05.
* 1°coefficient significantly different from one, p< 0.05.
ttBouyoucos and USDI were the same.
100
Slope = 0.995
80
R2 = 0.999
I 60
w
40
20
t
20
40
1
60
r-
20
100
80
oo
40
60
80
100
% Sand (Bouyoucos or USDI)
% Sand (Bouyoucos or USDI)
-
Slope * 1.044
R2 = 0.997
80
I
a0°^
1:1 line
D°jfl!
60
Bdr^
40
20
n
-
—i
20
40
60
1
80
100
% Sand (Qee and Bauder)
Fig. 2. Relationship between % sand measured by four methods.
sand content determined from hydrometer readings at 30 and
60 s, and readings at 40 s according to Bouyoucos (1962), was
within 0.5% by mass. They also found that the difference
between sand, as determined by sieve and hydrometer methods
often exceeded 5% by mass. Recently, Bohn and Gebhardt
(1989) stated that hydrometer readings anywhere between 30
and 60 s should reasonably estimate the sand content.
Relationships among the four methods of measuring clay
CANADIAN AGRICULTURAL ENGINEERING
content were all significant(Table II), but some ofthe relation
ships were quadratic insteadof linear and no pairof methods
gave identical results. As with sand content, this means that
each of the methods gave systematically different results from
the others, but the results from any method could be combined
with the regression equations to accurately predict the results
from any other method.
The relationshipbetween Bouyoucos (1962) and any of the
213
Table n. Comparison of percent clay content measured by four hydrometer methods.
Regression equation
Method
Y2= 1.079+0.735 (Yi) +
Gee and Bauder vs
SEof
SEof 1stdeg
SEof 2nddeg.
SEof
intercept
coefficient
coefficient
prediction
0.393t
0.021*
0.002#
0.992
1.39
0.671t
0.035*
0.004#
0.980
2.37
0.628t
0.033*
0.003#
0.982
1.39
0.001 (Yi)2
Bouyoucos
Y2= 1.362 +0.654 (Yi) +
Day vs Bouyoucos
0.003 (Yi)2
Y2= 1.098+0.712 (Yi) +
USDI vs Bouyoucos
0.002 (Yi)2
Day vs Gee and Bauder
Y2 = -1.818 + 1.056 (Yl)
0.257t
0.008*
0.991
1.59
USDI vs Gee & Bauder
Y2 = -0.845 + 1.035 (Yi)
0.219t
0.007*
0.993
1.36
USDI vs Day
Y2 = 1.004 + 0.977 (Yi)
0.125t
0.004*
0.998
0.81
t y-intercept significantly different from zero, p<0.05.
* 1°coefficient significantly different from one,p< 0.05.
# 2°coefficient significantly different from zero, p< 0.05.
other methods was quadratic (Fig. 3), and clay content deter
mined by the Bouyoucos (1962) method was always higher
than that determined by other methods. The other three meth
ods, (Day 1965), Gee and Bauder (1986), and USDI (1982)
gave comparable results. The 2 h reading in the Bouyoucos
(1962) method provided a mean estimate of particles about
0.0044 mm in diameter and less according to Stokes' law and
did not yielda correctestimateof the less than 0.002mmclay
ACKNOWLEDGEMENTS
We express our sincere appreciation to B. J. Sadasivaiah for
carrying out the laboratory analysis and to D. R. Bennett and
G. D. Buckland for their helpful suggestions. Funding for a
portion of this project was provided through the Summer
Temporary Employment Program sponsored by Alberta Ca
reer Development and Employment.
fraction.
Bohn and Gebhardt (1989) noted that clay content estimated
from 2 h readings was significantly differentfrom an average
estimate of the 6 and 12 h readings, and no statistical differ
ence was found between the 6 and 12 h methods. They
concluded that 6 h of settling should be adequate. Gee and
Bauder (1979) compared clay content determined on the basis
of graphical values (Day 1965) to clay content determined
from a formula calculation (Gee and Bauder 1979) and noted
no significant difference in the results.
CONCLUSIONS
Examination of four schedules of measurement by the hy
drometer method, for determination of PSD in soils from
southern Alberta, revealed statistically significant differences
in estimates for sand and clay content between methods.
These differences were most pronouncedin soils havinghigh
clay content. Conversion of sand and clay contents between
methods may be accurately completed using regression equa
tions presented in this study. Specification of the method used
for PSD measurements is essential because of the statistically
significant differences found between any of the four methods
compared.
214
KARKANIS, AU and SCHAAUE
Slop* = 0.888
80
-
/
R*» 0.980
X\ °
60
/*J&*
1:1 line
40
20
0
-
i
r-
1
1
20
1——i
40
1
1
60
1
80
100
% Clay (Bouyoucos)
Slope = 1.056
R2= 0.991
1:1 line
20
40
60
20
80
100
Slopes 1.035
80
100
Slope - 0.977
R2 = 0.993
80
| 60
60
1:1 line
40
60
100
80
I
40
% Clay (Gee and Bauder)
% Clay (Bouyoucos)
i
-
20
R2 = 0.998
-
^5°
1:1 line
jgr
40
20
T
20
40
60
80
1
1
20
100
1
1
40
T
1
60
1
1
80
1
100
% Clay (Day)
% Clay (Gee and Bauder)
Fig. 3. Relationship between % clay measured by four methods.
REFERENCES
ALBERTA AGRICULTURE. 1983. Standards for the classi
fication of land for irrigation in the Province of Alberta.
Alberta Agriculture, Lethbridge, AB.
BAVER, L. D., W.H. GARDNER and W.R. GARDNER.
1972. Soil physics, 4th ed. John Wiley and Sons Inc., New
York, NY.
BOHN, G. C. and K. GEBHARDT. 1989. Comparison of
hydrometer settling times in soil particle size analysis. J.
Range Manage. 42:81-83.
BOUYOUCOS, G. J. 1962. Hydrometer method improved for
making particle size analyses of soils. Agron. J. 54:464-465.
CANADIAN AGRICULTURAL ENGINEERING
DAY,P.R. 1965. Particle fractionation and particle-size anal
ysis. In C. A. Black, ed. Methods of soil analysis, Part 1.
Agronomy No. 9. Am. Soc. Agron., Madison, WI.
GEE, G. W. and J.W. BAUDER. 1979. Particle size analysis
by hydrometer: A simplified method for routine textural anal
ysis and a sensitivity test of measurement parameters. Soil
Sci. Soc. Am. J. 43:1004-1007.
GEE, G. W. and J.W. BAUDER. 1986. Particle size analysis.
In A. Klute, ed. Methods of soil analysis, Part 1. 2nd ed.
Agronomy No. 9. Am. Soc. Agron., Madison, WI.
USDI. 1982. Particle-size analyses. Chapter 4 of land classi
fication techniques and standards. Series 510. United States
Department of the Interior. Washington, DC.
215